UMI The source code of the SIGIR full paper "User-Aware Multi-Interest Learning for Candidate Mathching in Recommenders" in 2022.